deduplicate

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted prospect data from external sources like Apollo, Sales Navigator, and CSV files. * Ingestion points: External lead lists and contact records imported from Apollo, Sales Nav, and Clay. * Boundary markers: No explicit markers defined for data ingestion. * Capability inventory: Executes tool calls for person and company enrichment via MCP tools. * Sanitization: Mentions removing empty rows and flagging free email providers.
  • [EXTERNAL_DOWNLOADS]: The skill recommends executing the vendor's MCP tool via npx -y @coldiq/mcp@latest. This downloads and runs code from the author's package repository.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 08:56 AM
Security Audit — agent-trust-hub — deduplicate